Making third-party gauging data trustworthy between special surveys

The data is produced by someone you did not hire and cannot supervise

On a ship, hull thickness measurement is not performed by the owner's own inspectors. Class rules require the gauging to be carried out by a firm approved under the IACS procedural requirements for service suppliers, working to a survey plan the classification society issues before the special survey. The superintendent may never meet the operator who held the probe. The deliverable arrives as a class gauging report: a workbook of structural members, frame numbers and readings, signed by the firm and countersigned by the attending surveyor. That report is legally sufficient for the survey it was written for and almost useless as a dataset, because it was formatted for a surveyor reading one vessel at one drydocking, not for an engineer comparing the same longitudinal across three campaigns and two contractors.

This is what makes provenance the actual product of a thickness history module in marine work. When a reading is untraceable to a firm, an operator, an instrument and a method, the only thing the owner can defend is that class accepted it at the time. Retaining who produced each number is not administrative tidiness; it is the mechanism by which the owner can tell whether an apparent 0.8 mm step change between two campaigns is real wastage or the difference between two firms' conventions. Fleets discover this late, usually when a surveyor asks why the corrosion rate on a ballast tank longitudinal has reversed and nobody can answer without ordering staging.

A point on a hull is not a point on a drawing

In vessel and tank work the unit of continuity is a CML with a number. In marine work it is a structural member on a specific frame, on a specific side, in a specific compartment, at a specific height: No. 3 water ballast tank port, side shell longitudinal L14, frame 78, web. Every element of that string is somewhere two contractors will disagree. One writes Fr78, another FR-78, another gives the station in metres from the aft perpendicular. One calls the plate strake C, another counts strakes from the keel and calls it the third strake. One records the web, another the face plate, another averages the two and records the longitudinal.

The fix is not a better spreadsheet template. Contractors will not adopt it and class will not enforce it. The fix is a controlled structure catalogue built once from the vessel's midship section and structural drawings, in which every gaugeable member holds a stable identity for the life of the ship. Contractor files are mapped onto that catalogue at import, and the contractor's own label is retained as an alias, so the original class report stays reproducible while the trend is computed on a stable key. That alias table is what lets you open a 2016 campaign nine years later and know precisely which member was gauged.

Where the catalogue cannot resolve a point — a member absent from the drawings, or a reading taken on a repair patch — the import should force an explicit decision rather than quietly create a new location. Silent point creation is how a hull ends up with eleven thousand locations and no history.

Coating, probe and mode: why the same plate reads two thicknesses

Hull steel is painted, and often heavily. A conventional single-element or twin-crystal probe running in single-echo mode measures from the front surface of whatever it is coupled to, so a 300 to 500 micron coating system is read as steel. On a 12 mm side shell plate that is a four percent error, in the direction that flatters the owner. Echo-to-echo mode times between successive back-wall returns and ignores the coating entirely, but it needs a clean back wall, and heavy scale, lamination or a doubler will defeat it. Crews switch between modes as conditions demand and rarely say so in the report.

Velocity setting compounds the problem. A gauge calibrated on a steel reference block and then used on a member of different grade, or on structure at a very different temperature, drifts. Add couplant thickness, surface preparation quality, mill scale, and the possibility of an insert or doubler plate returning roughly double the expected wall, and the same physical member can plausibly return values a millimetre apart between two competent crews on two consecutive dockings.

This is why instrument and method are not metadata. The moment a step change appears in a trend, the first question is not whether the ship corroded but whether the method changed. If the record cannot answer that in seconds, the default action is a re-gauging campaign, which costs staging, tank entry and schedule at precisely the point in a docking where none of the three is available.

Averages, minima and the substantial corrosion threshold

Class rules do not treat every reading identically. Plating is generally assessed on an averaged value over a defined area, while isolated pitting and grooving are assessed against the single minimum found. Substantial corrosion — the condition that pulls a compartment into annual close-up survey and extended gauging in later surveys — is defined as wastage in excess of 75 percent of the allowable margin between as-built and renewal thickness, while still within acceptable limits. That is arithmetic performed on numbers whose convention has to be known before the arithmetic means anything.

The trap is comparing across conventions. Suppose a 2019 campaign reported a five-point average for a bottom plate and a 2024 campaign reported the lowest of five. The apparent loss is partly real wastage and partly the difference between a mean and a minimum. Run that across a whole ballast tank and you either manufacture substantial corrosion where none exists — buying annual close-up surveys for the remainder of the cycle — or, comparing the other way round, you average a genuinely pitted area into acceptability and carry a real defect forward.

A history module worth buying stores the convention on the reading, keeps the individual spot values behind any average, and refuses to compute a rate across mismatched conventions without flagging it. The point is not to prevent the comparison; it is to make sure the comparison is a decision someone took rather than a side effect of two spreadsheets being pasted together.

The points not taken are the survey finding

In most industries an unmeasured point is an inconvenience. In marine work it is a formal outcome. A gauging campaign is planned against a class survey plan that specifies required locations, and the reasons a location goes unmeasured are almost always operational: the tank was not gas-freed in time, staging was not erected, cargo remained aboard, the compartment could not be ventilated, the yard's entry permit did not issue, or the structure sits behind pipework or under a sounding pipe. The surveyor needs both the gap and the reason, because the gap becomes an outstanding item, a condition of class, or a re-gauging before the survey can close.

Storing exclusions as first-class records — location, planned method, reason not taken, who decided, what recovers it and when — converts a scatter of blanks into a work list the superintendent can act on before the next docking. It also protects the owner. Two years later, when a low reading turns up in an adjacent area, a record showing that this point was deliberately excluded for a documented access reason, with a recovery plan attached, is a very different position to defend than an empty cell nobody can explain.

The pattern of exclusions is itself diagnostic. If the same three ballast tanks go ungauged at every docking because gas-freeing runs late, that is a drydock planning defect rather than an inspection defect, and it is visible only if exclusions are queryable across campaigns instead of buried in the notes column of six different workbooks.

Steel renewal resets the clock, and most systems do not notice

Shipyards crop and renew plate. When a side shell strake or a group of longitudinals is renewed, the as-built thickness for those members changes, prior readings stop describing the metal now in place, and any corrosion rate computed across the renewal is meaningless. Systems that simply fit a line through every reading held for a location will report a hull regaining thickness, or a wastage rate near zero on a member that was renewed precisely because it was wasting fast.

Handling this properly means recording the renewal as an event against the structure identity, carrying its date, extent, the new as-built thickness and the specification of the new plate. History before the event is retained but excluded from rate calculations after it, and stays visible so a superintendent can see that this member has now been renewed twice in fifteen years. That second fact is frequently worth more than any individual reading, because a member renewed twice is telling you something about coating breakdown, drainage or ballast management that no thickness trend will ever say on its own.

What a technical superintendent should ask before buying

The evaluation questions that separate a real module from a document store are specific. Can it import a class-format gauging workbook without a consultant rewriting the mapping each time, and can it export back into that format for the surveyor? Does it hold a structure catalogue derived from the vessel's own drawings, with contractor aliases? Is measurement mode a controlled field or a free-text note? Can it record a point that was not taken, with a reason from a controlled list, and report those across a fleet? Does it treat steel renewal as an event? For any number on screen, can it show who took it, with what, when, and under which convention?

Then ask the operational questions. Can a superintendent with fifteen vessels see every member currently inside the substantial corrosion band across the fleet, sorted by next scheduled docking? Can ship's staff log coating condition and observed wastage against the same structure identity between dockings? Does the system tolerate a contractor who delivers a scanned PDF, or does it silently drop that campaign? The last question decides whether the dataset is complete in five years or has holes at exactly the campaigns that mattered most.

Atlantis builds this on an Odoo-based inspection management platform configured around vessel structure and the class survey cycle rather than a generic asset tree, and it is affordable, accessible and fully customisable to a fleet's own conventions. To see it run against one of your own gauging reports, request a demonstration at info@atlantisndt.com.

Request a consultation